I am writing my master thesis using pyfr to simulate the flow over a turbine cascade similar to T106c test case used in “F. D. Witherden and A. Jameson Impact of Number Representation for High-Order Implicit Large-Eddy Simulations” paper. However I have some questions about the setup and an error I am facing at:
The Error “Minimum sized time step rejected”, does it mean, that the simulation diverges? Sometimes the simulation goes on, when I just restart the simulation from the last solution, however sometimes I get the error at one specific time. I also tried to limit the time step using dt-max, but it did not work. If I turn off the PI controller, I get NaN values instead. I should mention that I have relatively fine mesh (comparing with your T106c test case). Do you have any suggestions, how I can prevent this error?
Do I need to use anti-ailiasing just for Polynomial orders higher than 2? Did you use it for the T106c case at P2 and Re=80000?
In my understanding the reference Reynolds number is not defined in pyfr in case of non-dimensionalaizing, therefore we have to eliminate it by choosing a specific reference dynamic viscosity, namely Mu_t0 = rho_t0 * a_t0 * Lc (Mu_t0 : Reference dynamic viscosity, rho_t0 : Reference density, Lc : Reference Lenght). So that the reference Reynolds-number would be 1. Is that correct?
1. The Error “Minimum sized time step rejected”, does it mean, that the simulation diverges? Sometimes the simulation goes on, when I just restart the simulation from the last solution, however sometimes I get the error at one specific time. I also tried to limit the time step using dt-max, but it did not work. If I turn off the PI controller, I get NaN values instead. I should mention that I have relatively fine mesh (comparing with your T106c test case). Do you have any suggestions, how I can prevent this error?
This typically means that the simulation is diverging. There are many possible reasons for this; including your mesh and start-up procedure.
2. Do I need to use anti-ailiasing just for Polynomial orders higher than 2? Did you use it for the T106c case at P2 and Re=80000?
Here it very much depends on the mesh, polynomial order, and Reynolds number. For this particular case I did not run with anti-aliasing, however I did need to be careful with how I started the simulation (running at p = 1 for a period of time before restarting at p = 2).
3. In my understanding the reference Reynolds number is not defined in pyfr in case of non-dimensionalaizing, therefore we have to eliminate it by choosing a specific reference dynamic viscosity, namely Mu_t0 = rho_t0 * a_t0 * Lc (Mu_t0 : Reference dynamic viscosity, rho_t0 : Reference density, Lc : Reference Lenght). So that the reference Reynolds-number would be 1. Is that correct?
The Reynolds number is typically fixed by changing the viscosity, yes.